AbstractIn this paper several simulations are presented of an

autonomous in-scale fast-ferry model TF-120 using data from a

remote Web-Wi-Fi platform for marine vehicles. The physical model
is developed to be autonomous and is controlled remotely from a PC
using Wi-Fi communications. An identification and validation of a
heading model is obtained with turning circle maneuverings on the
coastline of the Bay of Santander. Simulations of standard
maneuvering tests show the behavior of the identified model and the
kinematics. The parametric model identified is used to design
different classical control structures for heading autopilot. A
2-dimensional track-keeping system is developed based on line of
sight guidance scheme using the identified model of the autonomous
in-scale fast-ferry. The results of simulation are presented showing
the good performance of the guidance and control design.

Keywords Line of sight, parametric identification, heading
identification, heading control, simulation.
I. INTRODUCTION
physical in-scale model of the TF-120 turboferry (Fig. 2)
has been designed to be autonomous and is controlled
remotely from a Laptop in order to perform sea vessel
trials [13] with research purposes. This is part of a marine
vehicle Web-Wi-Fi platform [2], [11], [12] to carry out tests in
open waters that can not be made in a model basin and for at-
scale experimentation of coordination between sea vessels.
Autonomous guidance and control technologies are required to

Manuscript received June 3, 2008; Revised received September 21, 2008.
This work has been partially supported by MEC with the project DPI2006-
11835.
Francisco Jess Velasco is with the Dept. Electronic Technology &
Systems Engineering and Automatic Control, Univ. Cantabria, E.T.S. de
Nutica - C/ Gamazo 1 - 39004 Santander (Cantabria), Spain, (e-mail:
velascof@unican.es).
Elas Revestido is with the Dept. Electronic Technology & Systems
Engineering and Automatic Control, Univ. Cantabria, E.T.S. de Nutica - C/
Gamazo 1 - 39004 Santander (Cantabria), Spain, (e-mail:
revestidoe@unican.es).
Eloy Lpez is with the Dept. Sciences and Navigation Techniques,
Engines and Naval Architecture, Univ. Pas Vasco (UPV/EHU), E.T.S. de
Nutica y Mquinas Navales - C/ Ma Diaz de Haro 68 - 48920 Portugalete
(Vizcaya), SPAIN, (e-mail: eloy.lopez@ehu.es).
Emiliano Moyano is with the Dept. Applied Mathematics and Computing
Sciences, Univ. Cantabria, E.T.S. de Ingenieros Industriales y
Telecomunicaciones - Avda. de los Castros, s/n - 39005 Santander
(Cantabria) SPAIN, (e-mail: moyanoe@unican.es).
M. Haro Casado is with the Dept. Systems Engineering and Automatic
Control, Electronic Technology & Electronics, University of Cdiz Faculty of
Nautical Sciences - Polgono Rio San Pedro s/n, Edificio C.A.S.E.M. -
11510. Puerto Real. Cdiz, Spain.
perform these purposes. The present paper examines the
conventional line of sight (LOS) autopilot [14]. This kind of
autopilot directs the physical model along a course made up of
way points.
To be able to use LOS algorithm, it is necessary to develop
suitable heading controllers, meaning that the corresponding
parametric models are required.
In this work is developed a mathematical maneuvering
model, which includes the dynamics and kinematics, by means
of system identification techniques [6] using the data obtained
from turning circle maneuvers with a marine vehicle Web-Wi-
Fi platform for remote experimentation in the coastline of the
Bay of Santander.
Moreover, classical controllers are obtained for the
mathematical ship heading model which have been tested in
simulation. Genetic algorithm techniques have been used to
tune the controllers.
In this article simulations are presented in order to
demonstrate the performance of the guidance and control
design for TF-120 turboferry model. In subsequent research, a
track-keeping controller will be developed to be evaluated on
the coastline of the bay of Santander. Furthermore, the
possibilities of the platform for experimentation of marine
vehicles with research and educational purposes are discussed.


Simulations of an Autonomous In-scale
Fast-ferry Model
Francisco Jess Velasco, Elas Revestido, Eloy Lpez, Emiliano Moyano and M. Haro Casado
A

Fig. 1 Elements which make up the platform.
INTERNATIONAL JOURNAL OF SYSTEMS APPLICATIONS, ENGINEERING & DEVELOPMENT
Issue 3, Volume 2, 2008
114


II. PLATFORM ELEMENTS
Fig. 1 shows a schema of the on-board and on-land elements
which make up marine vehicle Web-Wi-Fi platform.
On-board Elements:
Components for the propulsion and steering of the vessel
(motors, servos, speed controller, turbojets).
Components of the control circuitry: PWM = Pulse Width
Modulation and receiver station.
Sensors (GPS, electronic gyrocompass, UMI and
accelerometers).
Communication components: industrial PC and access
point.
On-land Communication Elements:
One laptop with an access point.
One radio control station which takes over the steering of
the vessel in case of failure of the Wi-Fi network.
III. IDENTIFICATION
The variables that describe the movement of the physical
model on the horizontal plane (Fig. 2) are:
: Heading.
: Sway velocity.
u : Surge velocity.
U : Ship velocity.
: Turbojets angle.
The platform measures and stores in a file the heading data
obtained by the electronic gyrocompass which is used for the
identification.
A. Maneuver for Identification
For the identification of the heading, the turning circle
maneuver towards port is used (see Fig. 3). This maneuver
consists of two stages. In the first, the turbojet angle is at zero
and the physical model follows a constant heading. In the
second, the platform rotates the turbojet to +30 in one single
movement; then, the physical model begins to change its
heading towards port until it passes 540. The platform
maintains a constant speed throughout the whole maneuver
(u
0
= 0,4m/s). The wind speed is less than 0.5 m/s and the
current less than 2.5 m/s.
For the validation of the model, a turning circle maneuver
towards starboard is performed Fig. 4.

B. Identification of Heading Models
With the toolbox of Matlab ident [8], it is possible to
identify a model with different types of structures, including
continuous process models and first and second order Nomoto
models [4]. Nomoto models fit the following equations:

T K + = (1)

whose transfer function is:

( )
( )
1
K
s
s Ts

=
+
(2)

Second order Nomoto:



Fig. 2 Variables that describe the movement of the physical model
TF-120 on the horizontal plane.

0 5 10 15 20 25 30 35 40
-600
-400
-200
0
200

H
e
a
d
i
n
g
(

)
0 5 10 15 20 25 30 35 40
0
10
20
30
Time(s)
T
u
r
b
o
j
e
t

A
n
g
l
e
(

)

Fig. 3 Turning circle maneuver towards port.

0 10 20 30 40 50
0
200
400
600
800

H
e
a
d
i
n
g
(

)
0 10 20 30 40 50
-30
-20
-10
0
Time(s)
T
u
r
b
o
j
e
t

A
n
g
l
e
(

)

Fig.4 Turning circle maneuver towards starboard.
INTERNATIONAL JOURNAL OF SYSTEMS APPLICATIONS, ENGINEERING & DEVELOPMENT
Issue 3, Volume 2, 2008
115


1 2 1 2 3
( ) ( ) TT T T K T + + + = +

(3)

and whose transfer function is:

3
1 2
(1 )
( )
(1 )(1 )
K T s
s
s T s T s

+
=
+ +
(4)

Table I shows all the prediction-error model structures used
in the system identification process. There is also a continuous
process model structure.
Table II summarizes the results given by several model
structures. In all the cases, the percentage of output variations
reproduced by the model is calculated. A higher number means
a better model. The precise definition of the fit is:

hat
Fit [1 norm( ) / norm( mean( )]*100 y y y y = (5)

where y is the measured output and y
hat
is the
simulated/predicted model output.
The best model is the continuous model, which has a fit
value of 88.59%. Other structures were tested such as arx and
state space, but are not included in Table II as they have a
small value of fit.
The process of identification provides the following
coefficients for second order Nomoto:

1 2 3
0.94828, 2.3551, 0.5712, 1.3234 K T T T = = = = (6)

A 1st-order approximation is obtained by letting the
effective time constant be equal to:

1 2 3
0.94828, 4.2497 K T T T T = = + + = (7)

C. Validation of Heading Models
Fig. 5 shows how well the second order Nomoto model,
equation (4), fits the heading measured in the maneuver to
starboard.
A correlation analysis has been made on the prediction
errors. If there isnt autocorrelation of residuals for the output
heading, this means that the noise model structure is correct. In
the same way, if Cross correlation doesnt exist between the
residuals for the output (heading) and the input (turbojet angle)
confirms that the input/output model is correct.
Crosscorrelation between the residuals and the input values
appear adequate based on this validation (see Fig. 6).

TABLE I
MODEL STRUCTURES
Method Structures
Armax A(q)y(t) = B(q)u(t) + C(q)e(t)
Output error y(t) = [B(q)/F(q)]u(t) + e(t)
Box jenkins y(t) = [B(q)/F(q)]u(t) + [C(q)/D(q)]e(t)
Continuous Model (s) = G(s)(s)


TABLE II
RESULTS FROM THE IDENTIFICATION PROCESS
Type of model Coefficients Fit (%)
Continuous Model
K = -0.94828, T1 = 2.3551, T2 =
0.5712 T3 = -1.3234
88,59
Armax8221 A(q) = 1 - 0.5803 q^-1 - 1.051 q^-2
+ 0.4935 q^-3 + 0.001876 q^-4
- 0.06359 q^-5 + 0.008615 q^-6
+ 0.1124 q^-7 + 0.07884 q^-8
B(q) = 0.002071 q^-1 - 0.002289 q^-2
C(q) = 1 - 0.009262 q^-1 - 0.8128 q^-2
83,27
Output Error 221 B(q) = 0.005656 q^-1 - 0.00568 q^-2
F(q) = 1 - 1.998 q^-1 + 0.9983 q^-2
81,93
Box Jenkins 22221 B(q) = 0.005391 q^-1 - 0.005411 q^-2
C(q) = 1 + 0.7528 q^-1 - 0.09449 q^-2
D(q) = 1 - 0.05471 q^-1 - 0.9505 q^-2
F(q) = 1 - 1.999 q^-1 + 0.9986 q^-2

61,66


20 30 40 50 60 70
0
100
200
300
400
500
600
700
800
Time(s)
Estimated Heading()
Measured Heading()

Fig.5 Simulation of identified model with measured heading.

-20 -15 -10 -5 0 5 10 15 20
-0.05
0
0.05

Autocorrelation of residuals for output Heading
-20 -15 -10 -5 0 5 10 15 20
-0.05
0
0.05
Samples
Cross corr for input Turbojet angle and output Heading resids

Fig. 6 Prediction-error correlation analysis.
INTERNATIONAL JOURNAL OF SYSTEMS APPLICATIONS, ENGINEERING & DEVELOPMENT
Issue 3, Volume 2, 2008
116


IV. KINEMATICS OF THE IN-SCALE MODEL
In order to design a Track-keeping system, it is necessary to
calculate the actual position of the model TF-120. To define
the position of the vehicle, the following definitions are
required:
The vehicles flight path relative to the earth-fixed
coordinate system is given by a velocity transformation
according to Fig. 2:

1 1 2 1
( ) J v = (8)

where,

1 2 1
[ , , ] , [ , , ] , [ , , ]
T T T
x y z v u v w = = = (9)

1 2
( )
c c s c c s s s s c c s
J s c c c s s s c s s s c
s c s c c



+ + (
(
= + +
(
(

(10)

Assuming that in absence of the roll and pitch modes
( = = 0), the following equations are obtained:

cos sin x u v = +
(11)

sin cos y u v = (12)

r = (13)

V. HEADING CONTROL
A. Control Problem
An automatic pilot must fulfill two functions: course-
keeping and change of course. In the first case, the objective is
to maintain the trajectory of the vessel following the desired
heading ((t) = constant). In the second case, the objective is
to perform the change of heading without excessive
oscillations and in the minimum time possible. In both
situations, the correct functioning of the system must be
independent from the disturbances produced by the wind, the
waves and the currents.
The trajectory followed by a vessel can be specified by
means of a second order reference model:

( ) ( ) ( )
2 2
2
n n n r
t t t + + = (14)

where
n
is the natural frequency and ( 0.8 1), the
desired dampening coefficient system in a closed loop.
B. PID Controller Loop
With the first order Nomoto parameters obtained in (7), a
PID controller has been tuned to maintain the constant heading

d
.
For the correct functioning of the system, a dampening
coefficient in the range 0.8 1 is selected.
n
is selected as
the rotation angle of the turbojets

(rad./s) and the dynamics

of the physical model as 1/T (rad./s) according to the equation:

2 4 2
1
1 2 4 4 2
n
T

< + + < (15)

Next, a dampening value of = 0.9 is selected and the
following range of values fitted to
n
is obtained:

. .
0.1128 4.55
n
rad rad
s s
< < (16)

In keeping with equation (16), a value of
n
= 0.6 rad./s. can
be selected.
The schema of a conventional PID does not work very well
when the derivate mode amplifies the noise. To avoid this
problem, the following PID control schema is used [10]:

1
( ) 1
1
d
c
i d
T s
s K
E T s T s

(
= + +
(
+

(17)

Equation (17) shows that the derivate action is multiplied by
the term 1/(T
d
s + 1), which corresponds to a first order
system where T
d
is the time constant. The manufacturers give
as typical values of the range of 0.05 to 0.2, so = 0.1 is
selected.
Developing a procedure proposed in [4], the parameters for
the PID controller are obtained:

1.6133, 16.666, 1.3463
p i d
K T T = = = (18)

Since the results from the tuning, given in the following
sections, indicate that the integral action time is too high, a
first order network (19) has also been used in which the
integral action does not appear.

( )
s z
s K
E s p
( +
=
(
+

(19)

C. Tuning with Genetic Algorithms
A PID controller has been tuned by means of genetic
algorithms (Gas) [7]. The aim of the design is that the ship
should make a fast course change following, without
oscillations, the course determined by the values = 0.9 and

n
= 0.6 rad/sec in equation (14). The cost function selected
was:

1
)
n
i i
i
J(
=
= + (20)

Where is the vector of the controller parameters, n is the
INTERNATIONAL JOURNAL OF SYSTEMS APPLICATIONS, ENGINEERING & DEVELOPMENT
Issue 3, Volume 2, 2008
117


total number of iterations in the control system simulations,

i
the ith heading angle error between the desired and
obtained heading, is a scaling factor ( = 0.2 in this case)
and
i
. the ith rudder angle deflection. The term
i
has been
included in order to take into account also the minimization of
the control effort. A simulation time of 60 sec. has been used.
Each individual is represented by a parameter vector
= [K
p
T
i
T
d
] of the PID controller and = [k z p] for the first
order network. The chromosomes are of the binary type. The
selection of the range of values of the parameters was
performed with a view to avoiding an excessive saturation of
the actuators and to ensuring stable controllers.
The ranges of values of the parameters S selected were:

{0.1 10, 1 5000, 0.1 10}
p i d
S K T T = (21)

for the PID controller and

{0.1 100, 0.01 10, 0.1 20} S k z p = (22)

for the first order network.

For the optimization, a population of 30 individuals over
500 generations is used with a probability of crossover of 50%
and mutation of 5%. The Ga evaluates the cost function in
each iteration, after running the Simulink model, with the
controller. A roulette wheel was used in the selection, and the
principle of elitism was also used, keeping for the next
generation the best two individuals of the previous population,
and selecting them for the crossover and mutation.
The resulting parameters of the PID controller (17) are:

1, 21172, 4951.352255, 2.300027
p i d
K T T = = = (23)

And the resulting parameters the first order network in (19)
are:

32.603814, 0.42214, 8.696483 k z p = = = (24)

VI. 2-DIMENSIONAL LOS GUIDANCE SYSTEM
Systems for guidance are systems consisting of a waypoint
generator with human interface. One solution to design this
system is to store the selected way points in a way-point
database and use them to generate a trajectory (path) for the
ship Fig. 7. Other systems can be linked to this waypoint
guidance system as the case of weather routing, collision and
obstacle avoidance, mission planning, etc.
LOS schemes have been applied to surface ships by [9] and
[5]. In this methodology it is computed a LOS vector as the
vector from the body-fixed origin (x, y) to the next way-point
(x
k
, y
k
). This suggests that the set-point for the heading
autopilot should be chosen as:

1
( )
( ) tan
( )
k
d
k
y y t
t
x x t

| |
=
|

\
(25)

Where (x, y) is the vessel position usually measured with a
GPS. In this article, the position of the TF-120 model is
calculated with the kinematical equations (11) and (12) for
constant speed, so u = u
0
= 0.4m/s. Equation (25) requires a
sign test to ensure that
d
(t) is in the proper quadrant. The
autopilot follows the heading by guiding the TF-120 model
from way-point to way-point.
When moving along the path a switching mechanism for
selecting the next way point is needed. The way-point
(x
k+1
, y
k+1
) can be selected on a basis of whether the ship lies
within a circle of acceptance with radius R
0
around the way
point (x
k
, y
k
). Moreover if the vehicle positions (x(t), y(t)) at
time t satisfy:

2 2 2
0
[ ( )] [ ( )]
k k
x x t y y t R +
(26)

A guideline could be to choose R
0
equal to two ship lengths
(L
pp
), in the case of the in-scale physical model TF-120 model
L
pp
= 4.4m.

VII. COMPUTER SIMULATIONS
In the section below three types of simulations, made in
Simulink [3], are presented: maneuvering simulations, close
loop simulations and track-keeping simulations. The first ones
show the good performance of the model identified and the
kinematics (see section IV) for two typical maneuvers
performed on full-scale vessels. Close loop simulations point
out that the PID controller tuned in previous sections is
correct. The good result of these two kinds of simulations is
very important in order to obtain a good guidance system as
reflected by the last simulations.
A. Standard Maneuvers
The two figures below show simulations of turning circles
towards starboard for a rudder deflection of -30. The
evolution of the heading for a first order Nomoto model (2)
with the parameters in (7) is presented in Fig. 8. The trajectory
of the ship in Fig. 9 for a non-linear kinematical model


Fig. 7 Conventional LOS Guidance system.
INTERNATIONAL JOURNAL OF SYSTEMS APPLICATIONS, ENGINEERING & DEVELOPMENT
Issue 3, Volume 2, 2008
118


calculated in (11) and (12), with the only assumption that
v 0, describes a circumference which is typical of the turning
circle maneuver. This simulation verifies the suitability of the
model identified and the kinematical equations.
The zig-zag maneuver is performed in Fig. 10 as follows:
assuming that the simulation starts from an initial heading of
0, the simulation makes a change in the turbojet angle of 0 to
10. Thus, until the heading exceeds -10, the simulation does
not surpass from 10 to -10 of angle of the turbojet. In the
same way, until the platform exceeds 10 of heading, the
simulation does not change from -10 to 10 of angle of the
turbojet. A minimum of five cycles are required to perform the
full maneuver. Fig. 11 shows the trajectory of the ship for the
zig-zag maneuver
This maneuver establishes several important characteristics
of the yaw response. These are: the response time (time to
reach a given heading), the yaw overshoot (amount the vessel
exceeds 10 when the rudder has turned the other way), and
the total period for the 10 oscillations.
B. Close Loop Simulation
According to the heading control loop in Fig. 12, different
simulations are performed for the controllers tuned in previous
sections. In this block diagram characteristics of the servo as
an actuator are included, which are very important since they
can impose constrains on the control action (see Fig. 14). The
transfer function /
d
is the simplified model presented by [1]
0 10 20 30 40 50 60
400
200
600
800
1.000
1.200
1.400
1.600
-30
time(s)
A
n
g
l
e
(

)


Turbojet Angle
Heading

Fig. 8 Simulation of the Heading, with the Nomoto model (2) and the
parameters in (7) for a turning circle maneuver.
0 0.5 1 1.5 2 2.5 3
0
0.5
1
1.5
2
2.5
x(m)
y
(
m
)

Fig. 9 Trajectory of the ship with the model formed by the equations (2)
and non-linear equations (11) and (12) for a turning circle manoeuver.

0 5 10 15 20 25 30 35 40 45 50
-20
-15
-10
-5
0
5
10
15
20
time(s)
A
n
g
l
e
(

)


heading()
Turbojets Angle()

Fig. 10 Simulation of the heading for a 10 Zig-Zag maneuver.

0 2 4 6 8 10 12 14 16 18 20
-0.4
-0.2
0
0.2
x(m)
y
(
m
)

Fig. 11 Simulation of the position for a 10 zig-zag maneuver.
INTERNATIONAL JOURNAL OF SYSTEMS APPLICATIONS, ENGINEERING & DEVELOPMENT
Issue 3, Volume 2, 2008
119


and is the most commonly used in marine vehicles. This model
present two kinds of saturations:
Magnitude saturation: the turbojet motion is constrained to
move within a certain maximum angles. -
max
< <
max
,

max
= 30.
Slew rate saturation: the rate of turbojets is limited by a
maximum valued
max
as it was defined in equation (16).
A comparative study is made in Fig. 13 with different types
of controllers and tuning methods. PID Controllers are
represented with the parameters obtained in (18) and (23), and
also a first order network controller with the parameters in
(24). The graphical results given by the network controller
tuned with genetic algorithms are significantly much better
than with a PID controller tuned with classical methods (see
section V B).
The PID controller tuned with genetic algorithms gives a
response very close to the response given by the network
controller. This is because the integral action time is too high
in (23).
The control effort and the desired heading are plotted in
Fig. 14 for different control structures and tuning methods.
The control effort signal presents saturation at -30 or 30 due
to the limitations of the turbojet angle.
In Fig. 15 the trajectory of the ship is plotted for a first
order network controller which gives the best results.

C. Track-keeping Simulation
In this section is presented a conventional LOS guidance
system simulation. It has been used a first order network as a
heading controller with the parameters calculated in previous
sections. Fig. 17 shows the good performance of the LOS
algorithm and the heading controller (see Fig. 16) for a model
identified with the platform of marine vehicles and the
kinematical equations calculated in section IV.


Fig. 12 Heading autopilot block diagram including actuator saturations.

0 20 40 60 80 100 120
-30
-20
-10
0
10
20
30
time(s)
H
e
a
d
i
n
g
(

)


1st Order
Network()
Pid()
Pid Gas()
Desired
Heading()

Fig. 13 Heading autopilot for different control structures and tuning
methods.

0 5 10 15 20 25 30 35 40
0
1
2
3
4
x(m)
y
(
m
)
Fig. 15 Trajectory of the ship for a first order network controller tuned
with genetic algorithms

0 20 40 60 80 100 120
-30
-20
-10
0
10
20
30
time(s)
E
f
f
o
r
t
(

)


Pid Effort()
Pid Gas
Effort()
1st Order
Network
Effort()
Desired
Heading()

Fig. 14 Control efforts for different control structures and tuning
methods.
INTERNATIONAL JOURNAL OF SYSTEMS APPLICATIONS, ENGINEERING & DEVELOPMENT
Issue 3, Volume 2, 2008
120


VIII. CONCLUSIONS
Identification has been made of the heading model of the in-
scale TF-120 turboferry. With the tests carried out on the
coastline of the Bay of Santander, several prediction-error
model structures and continuous structures have been
identified with different orders. The best model has been
validated with a simulation on a data set different from the one
used for parameter estimation and with a correlation analysis
on the prediction errors. Simulations of standard maneuvers
show the response of the identified model and the kinematics.
The results from the tuning of the controllers by means of
Gas indicate that the integral action is too high, meaning that it
is advisable to use a first order network for the heading
controller.
The simulation has been carried out to verify the suitability
of the LOS algorithm, the heading controller and also the
identified model. An evaluation of a track-keeping controller
will be made on the coastline of the bay of Santander in further
research. Moreover, the possibilities of the platform for
experimentation of marine vehicles with research and
educational purposes are presented.
REFERENCES
[1] Van Amerongen J. Adaptive steering of ships- A model reference
approach to improved manoeuvering and economical course Keeping.
PhD thesis, Delft University and Technology, The Netherlands, 1982.
[2] F. J. Velasco, T. M. Rueda, E. Revestido, E. Moyano, E. Lpez and L.
A. Esquibel, Platform for remote experimentation of autonomous high
speed craft models with a combined technology for Wi-Fi and internet
communications programmed in LabVIEW, in Robotics and
Automation in the Maritime Industries, Chapter 12, Aranda, J. Gonzlez
de Santos, P. and De la Cruz, J.M., Ed. UCM, Madrid, Spain, 2006, pp
275293.
[3] Dabney, B. J., Harman, L. T., Mastering Simulink, Prentice hall, Upper
Saddler River, 2004.
[4] Fossen, T. I., Marine Control Systems: Guidance, Navigation and
Control of Ships, Rigs and Underwater Vehicles (Marine Cybernetics
AS). Trodheim, Norway, 2002.
[5] Fossen, T. I. Breverik, M., Skjetne, R., Line of sight path following of
underactuated marine craft. in: Proc. 6th IFAC Conf. on Maneuvering
and Control of Marine Crafts, Girona, Spain, 2003, pp. 244249.
[6] Ljung, L., System Identification Theory for the User, Prentice Hall, New
Jersey, 1999.
[7] Man, K.F.,Tang, K.S., Kwong, S., Genetics Algorithms: concepts and
designs. Springler, 1999.
[8] Matlab, System Identification Toolbox for Use with MATLAB, Userss
Guide, Version 6, 2005.
[9] McGookin, E.W., Murray-Smith, D.J., Lin, Y., Fossen, T.I., Ship
steering control system optimization using genetic algorithms, Journal
of Control Engineering Practice, Aug. 1998, pp. 429443
[10] Smith, C. A., Corripio, A. B., Principles and Practice of Automatic
Process Control, John Wiley & Sons, Canad, 1997.
[11] Velasco, F. J., Rueda, T. M., Revestido, E., Esquibel, L. ., Marine
Vehicles Web-Wi-Fi Platform for Remote Experimentation Programmed
in LabVIEW, in 'ACE 06' 7th IFAC Symposium on Advances in
Control Education, Madrid, 2006.
[12] Velasco, F. J., Rueda, T. M., Revestido, E., Esquibel, L. ., Platform
of Trials in an Open Environment for the Cooperation of Marine
Vehicles Programmed in Labview with Wireless Connectivity by Means
of Wifi, in WSEAS/IASME Int. Conf. on Electroscience and
Technology for Naval Engineering and All-Electric Ship, Miami, 2005.
[13] Velasco, F. J., Rueda, T. M., Revestido, E., Esquibel, L. ., Turning
Circles of an Autonomus High Speed Craft Model, in 5th WSEAS Int.
Conf. on System Science and Simulation in Engineering. 1st WSEAS
International Symposium on Advances in Naval Science, Research and
Education. Tenerife, Spain, 2006.
[14] Velasco, F. J., Revestido, E., Lpez E., Moyano E., Haro Casado M.,
Autopilot and Track-Keeping Simulation of an Autonomous In-scale
Fast-ferry Model, in 12th WSEAS International Conference on
SYSTEMS. Heraklion- Crete, Greece, 2008.
0 50 100 150 200 250 300 350
0
20
40
60
80
100
120
140
160
180
time(s)
H
e
a
d
i
n
g
(

)


Heading()
Desired Heding()

Fig. 16 Heading control of the track-keeping system using equation (26)
as a switching mechanism.
0 10 20 30 40 50 60 70 80
-30
-25
-20
-15
-10
-5
0
5
10
15
20
y(m)
x
(
m
)
Simulated trajectory
Way-points

Fig. 17 xy plot of the simulated trajectory of the TF-120 model and
desired geometrical path made up of way-points.
INTERNATIONAL JOURNAL OF SYSTEMS APPLICATIONS, ENGINEERING & DEVELOPMENT
Issue 3, Volume 2, 2008
121